Better support for modeling recreational discards

[iantaylor-NOAA]

Three general approaches to recreational discards are as follows. These are similar to the approaches used for commercial discards but with some additional quirks (e.g. death-by-depth matrix). It would be good to both better document these options and provide more guidance on the reasons to choose among them. It would also be good to develop some generalized code to support the more complex ones.

1. add dead rec discards to retained catch
   pros: easy, allows model to account for total mortality, probably makes sense when dead discards are small compared to retained catch
   cons: assumes discards are from same subset of population as retained catch, no clear way to benefit from discard length data
2. model rec discards as a separate fleet
   pros: flexible, makes use of length data, only requires modeling dead discards
   cons: requires extra assumptions to extrapolate outside the range of years with discard estimates, requires more selectivity parameters (presumably dome-shaped), ratio of discard to retained doesn't influence model (could be a con if you don't think this information is useful), selectivity for both discard and retained fleets may need to be time-varying to account for changes in size limits
3. model rec discards via a retention function
   pros: provides estimates of discards outside the range of data, time-varying retention can account for changes in size limits
   cons: requires modeling all discards and then applying a discard mortality rate, the death-by-depth matrix used to calculate recreational discard mortality means that the discard mortality will be different each year, SS doesn't currently support separate data weights for retained vs. discard within a single fleet